Evaluation of the thyroid nodule

Oral Oncology 49 (2013) 645–653

Contents lists available at ScienceDirect

Oral Oncology journal homepage: www.elsevier.com/locate/oraloncology

Evaluation of the thyroid nodule Bryan McIver Moffitt Cancer Center, 12902 Magnolia Drive, Tampa, FL 33612, United States

a r t i c l e

i n f o

Article history: Available online 23 May 2013 Keywords: Thyroid nodule Molecular marker Molecular cytology Fine needle biopsy Thyroid cytology

s u m m a r y Thyroid nodules are the most common endocrine tumor, affecting over 50% of women over age 50. While the incidence of thyroid cancer is increasing, most thyroid nodules ultimately prove benign. An accurate, costeffective algorithm for the evaluation of thyroid nodules includes clinical, biochemical, ultrasonographic and cytopathologic evaluation. The development of molecular diagnostics promises to transform the interpretation of previously indeterminate cytology results, but application of these techniques requires a thorough understanding of the impact of pre-test probability of malignancy on the performance of the molecular assays. Efforts to avoid diagnostic surgery in nodules that are cytologically indeterminate are admirable, but may be premature, pending additional confirmatory studies that prove this approach is safe. Meanwhile, the finding of oncogene mutations has sufficient specificity for cancer that this approach might be useful to tailor surgery. However, the cost effectiveness of this approach requires further evaluation. Ó 2013 Elsevier Ltd. All rights reserved.

Introduction Thyroid nodules are the most common endocrine tumor, with population-based screening studies identifying clinically palpable nodules in 5% of adults, while ultrasound and autopsy studies demonstrate nodules in more than 50% of women and 20% of men over age 50.1 Thyroid cancer, which usually presents as a nodule, is increasing in incidence more rapidly than any other cancer type.2 However, fewer than 5% of nodules, whether palpable or incidentally discovered, prove to be malignant in autopsy series.3 Although selection of nodules for biopsy based on suspicious ultrasound features enriches the yield of malignant nodules, the proportion that prove malignant remains a mere 10–15%.4 The primary challenge in the evaluation of thyroid nodules is to reliably identify the majority of benign nodules that do not require surgical removal, while avoiding the risk that malignant (or pre-malignant) nodules are not identified, so missing an opportunity to provide effective early surgical treatment. More than 450,000 thyroid biopsies and over 120,000 thyroid surgeries are performed each year in the US, but with an estimated 72 million adults living with a thyroid nodule, it is likely that at least four times as many nodules develop than are currently being evaluated each year. The number of FNA procedures being performed continues to rise at a rapid pace and may be partly responsible for the increasing incidence in diagnosed thyroid cancer.5 With the increasing use of imaging technologies that identify thyroid nodules, including the use of thyroid ultrasound as a screening tool in some settings, the cost and morbidity of thyroid nodule

E-mail address: Bryan.McIver@moffitt.org 1368-8375/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.oraloncology.2013.03.435

evaluation threatens to become a significant public health expenditure issue, so an efficient, cost-effective strategy for the evaluation of thyroid nodules is an important goal of current and future clinical practice. Pathogenesis of thyroid nodules The current approach to thyroid nodule evaluation is to categorize nodules as either ‘‘benign’’, ‘‘indeterminate/suspicious’’ or ‘‘malignant’’. However, this classification encompasses a broad range of pathologic processes, including several types of malignancy; follicular adenomas (neoplastic), some of which might be pre-malignant; adenomatous nodules (non-neoplastic); pseudonodules within a multi-nodular gland (hyperplastic); simple thyroid (colloid) cysts; and regenerative nodules within a gland involved by Hashimoto’s thyroiditis.6 In developing the appropriate diagnostic and management strategy for any thyroid nodule, it is important to establish, as far as possible, the underlying pathogenesis, something that requires a comprehensive assessment, including clinical assessment, laboratory evaluation, ultrasound examination and, in many cases, cytologic evaluation of fine needle aspiration material (FNA). Increasingly, attention is also being paid to genetic and other molecular markers, as an adjunct to thyroid cytology, though the role of these markers is not yet fully established or accepted in current published guidelines.7–9 Follicular adenoma Activating mutations of the thyrotropin (TSH) receptor have been described in a proportion of these nodules, while mutations

646

B. McIver / Oral Oncology 49 (2013) 645–653

of RAS and chromosomal rearrangements involving the PPARc gene, including the relatively common PAX8/PPARc rearrangement, are present in others.10 A proportion of these adenomas, particularly those with activating mutations of the TSH receptor, are functionally active tumors, producing and releasing thyroid hormone autonomously and ultimately leading to TSH suppression and thyrotoxicosis. So-called hot nodules (those that produce enough thyroid hormone to cause at least biochemical hyperthyroidism) are rarely malignant and are usually treated with radioactive iodine. However, a proportion of ‘‘cold’’ follicular adenomas, even though technically benign may be pre-malignant and therefore warrant excision in their own right.11 Because FNA cytology cannot distinguish follicular adenoma from carcinoma, the recommended management of these nodules has been surgical resection, so the issue of pre-malignant follicular neoplasms has been a theoretical rather than a practical consideration. However, several recent studies suggest that genetic markers or gene expression profiling could make it possible to distinguish benign from malignant follicular neoplasms.12,13 These studies have already led the National Cancer Cooperative Network (NCCN) to recommend observation as an alternative to surgery for these nodules, if the ‘‘molecular profile suggests a risk of malignancy of <5%’’.7 Although it has been a long-held goal in the thyroid community to reduce the number of ‘‘unnecessary’’ thyroid surgeries for indeterminate cytology, these approaches do create some risk that that pre-malignant lesions that appear molecularly benign could be left in situ. This remains a challenge to the early adoption of such assays and will require careful long-term follow-up studies of those patients whose follicular adenomas are not recommended for surgery. Before the widespread adoption of this technology, a formal registry study should be designed and implemented to ensure these patients are not lost to follow up after a benign molecular diagnosis. Thyroid cysts Pure cysts exhibit the histologic features of a simple cyst, with proteinacious cyst fluid surrounded by a simple cuboidal epithelium.14 The cyst material closely resembles the colloid within a normal thyroid follicle. Other cysts form in the context of preexisting, usually benign, thyroid nodules, in which central degeneration arises as a result of an inadequate blood supply, or, less commonly, hemorrhage following trauma, or spontaneous intralesional bleeding. Less often, a cyst forms in response to a thyroid carcinoma, most often papillary carcinoma, which may represent only a small solid nodule on the wall of the cyst.15 Distinguishing a benign from a malignant cyst can sometimes be challenging, because cyst fluid contamination may yield non-diagnostic FNA cytology. Ultrasound features are crucial in assessing these cysts, since true simple cysts are almost never malignant, while the solid component of a cyst wall may be the indication of a malignant lesion. Pure cysts often reaccumulate fluid following even near-complete aspiration, so large cysts may require surgical resection. Ablative procedures, including ethanol injection following cyst aspiration, are also available, but have not been widely accepted in the US, though they are more popular in some European and Asian countries.16 Multinodular goiter The cause of multinodular goiter is unknown, though iodine deficiency in childhood and a genetic predisposition have been implicated in the later development of this condition.17 The natural history of a multi-nodular goiter is slow progressive growth over years or decades, with gradually developing autonomy and eventual subclinical hyperthyroidism, or clinical thyrotoxicosis. There is some evidence of abnormal regulation of the second messenger

system that leads to thyroid follicular cell growth, and a proportion of these goiters arise in response to polymorphisms of the TSH receptor.17 The risk of malignancy within multinodular goiter has variously been reported to be lower or higher than that within a solitary nodule, but a recent meta-analysis suggests that thyroid cancer is likely less frequent in multinodular goiters than in solitary nodules, particularly outside the US and particularly in iodine deficient areas.18 However, the possibility of malignancy within a multinodular goiter must still be considered and evaluation undertaken accordingly. Benign adenomatous nodule These nodules are non-clonal, arising from a group of follicular cells rather than from a single cell that has undergone neoplastic transformation.19 The stimulus for the growth of such a nodule is unclear, but does not seem to arise from abnormalities of the TSH-receptor or G-protein system. In many cases, the serum TSH concentration in these patients is normal and the follicular cells that make up the nodule are likely sensitive to TSH, raising the possibility that thyroid hormone suppressive therapy could be effective to stabilize or shrink such nodules.20 However, this approach can only safely be undertaken if the benign nature of the nodule is firmly established beforehand. Regenerative nodule Perhaps the most common cause of thyroid nodules, particularly in young women, is Hashimoto’s disease.21 In response to the inflammatory process and chronic destruction of the parenchyma, the thyroid’s capacity to secrete thyroid hormone is progressively reduced. TSH begins to rise, and there is increasing evidence that a TSH concentration even in the upper normal range may represent early thyroid failure among patients with Hashimoto’s disease.21 TSH stimulates growth of the remaining viable thyroid cells. However, it is not always possible for the entire gland to respond, particularly in the presence of intra-glandular fibrosis. A situation analogous to that of hepatic cirrhosis arises. In most cases, a ‘‘micronodular’’ pattern develops, creating the classic clinical features of a ‘‘bosselated’’ goiter. In some cases however, a ‘‘macronodular’’ pattern develops, with the development of a discrete nodule within the gland. Sometimes these nodules can grow rapidly, under the regulation of TSH or of TSH-receptor antibodies. It is for this group of patients that thyroid hormone suppression is most likely to be of benefit, and a trial of thyroxine treatment is certainly justified in this context, once the benign nature of the nodule has been reliably confirmed. Thyroid cancer Almost all thyroid cancers present initially as a thyroid nodule, either with or without neck node involvement at the time of diagnosis. The various types of thyroid cancer and their management are discussed elsewhere in this volume, but the goal of thyroid nodule evaluation is to reliably identify as close to 100% of thyroid cancers as possible, while avoiding the resection of asymptomatic, truly benign thyroid nodules. Evaluation of thyroid nodules Several professional organizations, including the American Association of Clinical Endocrinologists (AACE), the American Thyroid Association (ATA) and NCCN have published guidelines for the evaluation of thyroid nodules, all of which endorse a similar multistep strategy: clinical assessment, measurement of TSH (to identify

B. McIver / Oral Oncology 49 (2013) 645–653

647

Figure 1. Recommended size thresholds for consideration of thyroid FNA as part of the evaluation of a thyroid nodule, from the American Thyroid Association Guidelines for the Management of Thyroid Nodules and Thyroid Cancer.8

hot nodules), ultrasound evaluation, and biopsy of nodules selected according to size and ultrasound characteristics (Fig. 1).7–9 The primary goal of such an assessment is to distinguish reliably the malignant (and pre-malignant) minority from the benign majority of nodules. It is important to recognize, however, that no such evaluation can ever be entirely fool-proof, so some small proportion of cancers will be missed by any evaluation (false-negative results), while a number of entirely benign nodules will be recommended for surgical resection, because of our inability to clearly define those nodules as benign (false-positive results). The goal of any evaluation algorithm is to maximize both its negative predictive value (NPV), so that surgery can be safely avoided for benign nodules, and its positive predictive value (PPV), so that cancers can be effectively treated at the first surgical procedure. Clinical evaluation Clinical assessment remains a crucial first step in the evaluation of a thyroid nodule.8,9 The history should cover the duration and growth pattern of the nodule; previous thyroid diagnoses or evaluations; local compressive symptoms; invasive symptoms; systemic symptoms that suggest thyrotoxicosis or hypothyroidism; personal history of head or neck irradiation; family history of thyroid cancer, or of benign thyroid disease; and possible cancer predisposition syndromes, particularly MEN-2 or Cowden’s syndrome. The examination should assess the thyroid and the nodule, the loco-regional lymph nodes, voice, swallowing mechanism and the state of the thoracic inlet (Pemberton’s sign), as well as the patient’s clinical thyroid status. Large, compressive nodules, causing local symptoms (dysphagia, dysphonia, dyspnea) are likely to require surgery, even if benign, so the evaluation should be targeted to defining the extent of surgery needed. While a clear diagnosis is always desirable pre-operatively, the application of expensive molecular markers, for example, to define the benign nature of such a nodule is of

questionable utility in this setting. Consequently, it is the PPV of the evaluation that matters most, so that cancer can be accurately diagnosed pre-operatively and the extent of surgery determined accordingly. The NPV, in this setting is less relevant, because the intent will not be to avoid surgery. Similarly, there are several high-risk situations in which the goal should not be to avoid surgery: Patients with a history of radiation exposure in childhood, or a strong family history of medullary thyroid cancer, who now present with enlarging thyroid nodules may be appropriate to consider surgical candidates, even if the nodule being evaluated proves to be benign. It is less clear to what extent a family history of differentiated thyroid cancer should be regarded as a risk factor, but a proportion of these malignancies are familial, so the pre-test probability of malignancy in that setting is higher than for the population as a whole. Similarly, men with large, solitary nodules are more likely than average to harbor malignancy. It goes almost without saying that the presence of vocal cord paralysis, hemoptysis, significant neck pain (especially radiating ipsilaterally to the ear or throat), upper extremity neurological symptoms, lymph node enlargement, or rapidly enlarging thyroid masses should all be regarded as high risk features and handled accordingly. Subsequent findings that suggest benign disease should be treated with skepticism and additional confirmatory studies should be considered in the appropriate clinical setting. Laboratory evaluation For asymptomatic nodules, each of the guidelines suggests measurement of serum TSH as the initial diagnostic step. Suppressed TSH should lead to consideration of a hot nodule, which may be appropriate to consider for non-surgical treatment, usually radioactive iodine. Malignancy in this setting is rare, though not unknown, but the low true positive rate and the high false positive rate for FNA cytology in a hyperfunctioning nodule make FNA less useful for these nodules. The finding of a suppressed TSH leads to a

648

B. McIver / Oral Oncology 49 (2013) 645–653

recommendation to undertake isotopic imaging of the thyroid, proceeding with FNA only for cold nodules.7–9 The finding of a normal or elevated TSH mandates further evaluation of the nodule to exclude malignancy. It is worth noting that TSH concentrations are correlated with the risk for malignancy, even for TSH towards the upper end of the normal range, a finding that remains incompletely understood, but is of little practical utility, due to a lack of any TSH concentration that provides clear demarcation between benign and malignant nodules.22 Measurement of serum calcitonin has been advocated by some as an initial step in thyroid nodule evaluation, and is supported by guidelines from the European Thyroid Association.23 The costeffectiveness of this approach has been questioned, given the low incidence of medullary thyroid carcinoma (MTC), but there is little doubt that MTC can be hard to diagnose on FNA cytology, and widespread agreement that a pre-operative diagnosis of MTC, rather than the common diagnosis of ‘‘suspicious’’ or ‘‘follicular neoplasm’’ can be helpful in guiding the extent of initial surgery. Consequently, it seems reasonable to consider measurement of serum calcitonin in patients with thyroid nodules that fall into these particular cytologic categories. Ultrasound examination and other imaging Real-time, 2-dimensional, high resolution ultrasound has become the widely accepted standard of care for the evaluation of thyroid nodules, because of its measurement accuracy, ability to characterize nodules, capacity to identify subclinical nodules, and utility in characterization and mapping of pathological lymph nodes. All thyroid nodules undergoing evaluation, except those hot nodules noted above, should be evaluated ultrasonographically.8,9 For maximum effectiveness, ultrasound should be used as a real-time imaging technique, which is most effectively performed by the treating physician, because the dynamic findings may not be fully captured in static images, while the accuracy of FNA is clearly improved by the use of ultrasound.24 Considerable effort has been expended, by a number of groups, to define the ultrasound characteristics that best distinguish benign from malignant thyroid nodules. Each of the features of malignancy has reasonable specificity for cancer (averaging 60–90%, Table 1). However, sensitivity is less impressive, which results in a substantial risk that a malignant nodule could be missed if the decision to perform FNA is based purely on the presence of these features of malignancy.25 Furthermore, most benign nodules also exhibit at least one of the so-called ‘‘features of malignancy’’, most often hypoechoic echo texture. A combination of several malignant features certainly increases the probability of the nodule proving malignant on FNA cytology, however, particularly when demographic features are also taken into account, with a PPV in excess of 90% in these cases; Similarly, the complete absence of those malignant features, in a nodule that is spongiform with internal cystic spaces, results in a NPV that also exceeds 90%.26 Unfortunately, most nodules lie at neither end of this spectrum, so the majority of nodules are at least candidates for FNA.

In an effort to avoid the temptation to biopsy every nodule the ATA Guidelines developed a relatively complex algorithm to determine which nodules are worthy of biopsy.8 As shown in Fig. 1, these criteria are based on both the size and ultrasound features of the nodule, but also include an assessment of risk from clinical assessment. ‘‘High risk’’ patients are deemed to include those with a history of radiation exposure, positive family history, or previous thyroid malignancy. In these cases, nodules with one or more suspicious ultrasound features should be considered even when they are as small as 5 mm. In contrast, in a low risk patient, with classic benign features, the recommendation is made to proceed with biopsy only when the nodule exceeds 2 cm in maximum dimension. Several advanced ultrasound techniques have been proposed to improve the accuracy of traditional ultrasound in the evaluation of thyroid nodules. Assessment of blood flow, using Doppler techniques has become accepted as a useful tool and is incorporated into the ATA guidelines. Elastography, a technique that assesses the stiffness of a nodule through measurement of ultrasonographic compressibility has shown promise in highly specialized hands, and seems likely to gain wider acceptance once the problem of inter-operator variability has been improved.27 Alternative imaging modalities, including CT and MRI, can be helpful to more accurately define the neck anatomy, particularly ahead of thyroid surgery for large nodules or invasive cancer, but play little role in the initial evaluation of most thyroid nodules.8,9 In contrast, positron emission tomography (PET) has been studied as a possible means to distinguish benign from malignant nodules.28 Many thyroid cancers, especially of the less well differentiated types, are PET-positive, while most benign nodules are PETnegative. However, neither the discriminatory power of this technique, nor the cost is sufficiently favorable to support routine use of PET scanning in the initial assessment of a thyroid nodule.28 Ultrasound is also useful in the guidance of a fine needle aspiration, to ensure the appropriate nodule is identified, to guide the needle to its target and to document that the aspiration is performed while the needle remains in the appropriate location.29 Using ultrasound guidance improves diagnostic yield and reduces the number of indeterminate cytology calls, probably by increasing the yield of follicular cells, while allowing finer gauge needles to be used, which reduces red blood cell contamination of the smears.30 Additional benefits accrue from the ability to target specific locations within the nodule – the solid nodule on the wall of an otherwise cystic lesion, for example – and to assist in draining thyroid cysts to complete dryness. Ultrasound evaluation and ultrasound guidance for FNA should be considered ‘‘standard of care’’ for the evaluation of thyroid nodules. FNA cytology The introduction of FNA cytology more than three decades ago transformed the approach to thyroid nodules and has been an unquestioned success. Prior to its use, almost all thyroid nodules

Table 1 US characteristics of malignant thyroid nodules, with associated sensitivity, specificity, positive predictive value and negative predictive value. From: Frates MC, Benson CB, Charboneau JW. Management of thyroid nodules detected at US: society of radiologists in ultrasound consensus conference statement. Radiology 2005;237:794–800. US feature

Sensitivity (%)

Specificity (%)

Positive predictive value (%)

Negative predictive value (%)

Microcalcifications (1–5) Hypoechogenicity (2–5) Irregular margins or no halo (2–5) Solid (4–6) Intranodule vascularity (3, 6) More tall than wide (2)

26.1–59.1 26.5–87.1 17.4–77.5 69.0–75.0 54.3–74.2 32.7

85.8–95,0 43.4–94.3 38.9–85.0 52.5–55.9 78.6–80.8 92,5

24.3–70.7 11,4–68.4 9.3–60.0 15,6–27.0 24.0–41.9 66.7

41.8–94.2 73.5–93.8 38.9–97.8 88.0–92.1 85.7–97.4 74.8

649

B. McIver / Oral Oncology 49 (2013) 645–653 Table 2 The Bethesda Classification for thyroid cytopathology reporting. From Cibas and Ali.32 Diagnostic category

Risk of malignancy

Usual management

Nondiagnostic or unsatisfactory Benign Atypia of undetermined significance or follicular lesion of undetermined significance Follicular neoplasm or suspicious for a follicular neoplasm Suspicious for malignancy Malignant

1–4 0–3 5 – 15 15 – 30 60 – 75 97 – 99

Repeat FNA with ultrasound guidance Clinical follow-up Repeat FNA Surgical lobectomy Near-total thyroidectomy or surgical lobectomy Near-total thyroidectomy

were excised because of the small, but finite, risk of malignancy. In the Mayo Clinic experience between 1979 and 1990, the use of FNA cytology reduced the proportion of surgeries for what proved to be benign thyroid nodules by 50%, while doubling the likelihood of detecting malignancy, an experience mirrored in other practices.31 It is crucial to remember, however, that, except in the case of papillary thyroid cancer (PTC), with its characteristic nuclear features, thyroid cytology alone rarely renders a diagnosis. Rather, the goal is to provide reassurance that the risk of malignancy is sufficiently low that observation is appropriate in the right clinical setting, or that the risk of malignancy is high enough that surgery is justifiable. It is, in fact, an exercise in probabilities. Nevertheless, when the needle aspirate contains sufficient material for adequate assessment, cytology yields an accurate and reliable diagnosis of a ‘benign nodule’ in most cases (range 62–85%), and the majority of these patients can be safely observed. Approximately 5% (range 1–8%) of nodules are cytologically malignant (typically PTC). However, 15–30% of biopsied nodules exhibit ‘‘indeterminate’’ cytology, including several categories of follicular lesions, with various probabilities of malignancy.32 In an effort to standardize the terminology, particularly for indeterminate nodules, and to more accurately define the post-test risk of malignancy, Cibas et al. in 2009, convened an expert working group that developed the ‘‘Bethesda Classification’’ for thyroid cytology (Table 2), a classification that closely resembles others in use Worldwide.33 The estimated risk of malignancy, according to that expert panel, within each of the six cytological classes is shown in the table, and range from <5% for ‘‘benign’’ and recurrently ‘‘non-diagnostic’’ categories, to >95% for the ‘‘malignant’’ category. Intermediate categories are expected to have an intermediate risk of malignancy, with ‘‘Atypia of Uncertain Significance’’ and ‘‘Follicular Lesion of Uncertain Significance’’ (AUS/FLUS) having 5–15% probability of cancer, while ‘‘Follicular Neoplasm’’ (FN) has 15–30% malignancy risk. In reality, however, the variability in risk of malignancy (ROM) in each of these categories is much broader than this consensus panel would suggest. As demonstrated by Wang et al. in a formal meta-analysis of the literature, the risk of malignancy in nodules

classified as cytologically benign has been reported in several large series to range from 2% to 18%, with an average of 6% (Table 3), while cytologically malignant nodules prove to be cancer between 95% and 100% of the time.34 A rather greater discrepancy exists in the ROM of nodules whose cytology is read as FN, which carries a malignancy risk that ranges from 14% at Mayo Clinic to 49% at Washington Hospital, two centers that have reported the largest series of such FNA cytology. The variability in malignancy risk is even greater in the Bethesda Category 3 (AUS/FLUS), ranging from 0% at Abington Hospital in Pennsylvania, to 48% at both Yale University and Ball Memorial Hospital. These broad discrepancies in ROM between centers of excellence do not seem likely to reflect differences in patient populations, nor ‘‘inexperience’’ of the cytopathologists at some of the largest thyroid centers in the country. Instead, it is probable that these high volume centers simply take different approaches to assigning Bethesda groups. It is beyond the remit of this article to discuss the details of cytopathologic classification, but it is instructive to review the criteria by which a cytologic sample can be included in the AUS/FLUS category: Samples that exhibit ‘‘air drying artifact’’ and ‘‘clotting artifact’’ (which presumably do not alter the probability of malignancy in the lesion); ‘‘more prominent than usual population of microfollicles’’ (which might increase the ROM towards that of FN); or ‘‘focal features suggestive of papillary carcinoma’’ (which would significantly increase the risk of this sample ultimately proving to be malignant) all can be – but need not be – reported as AUS/FLUS. Inclusion is at the discretion of the cytopathologist, so that, at Mayo Clinic for example, the presence of ‘‘focal features suggestive of papillary carcinoma’’ would probably place that sample into the ‘‘Suspicious for malignancy’’ group, rather than AUS/FLUS, while significant clotting or drying artifact would likely be designated as non-diagnostic (personal communication: Dr. Thomas Sebo, 2012). These differences reflect a fundamental limitation of cytology. However, until recently, this was of little direct relevance, because the only crucial diagnosis was that of a benign nodule, which made it unlikely the patient would undergo surgery. All other diagnostic categories led to surgical resection in most cases, making a histo-

Table 3 Meta-review of thyroid cytology studies, published since 2001, during the era of near-universal ultrasound guidance. From: Wang et al.34 Study

Blansfield et al. (20) (Abington Hospital, PA) Sclabas et al. (21) (MD Anderson) Castro and Gharib (22) (Mayo Clinic) Wu et al. (23) (Ball Memorial, IN) Yassa et al. (9) (Brigham & Women’s) Yang et al. (24) (Northshore LJJ & U. Texas Galveston) Oertel et al. (25) (Washington Hospital) Banks et al. (26) (Johns Hopkins) Nayar and Ivanovic (27) (Northwestern) Theoharis et al. (28) (Yale) Faquin and Baloch (29) (MGH &. University Pennsylvania) Summary

No. resected in series

183 240 1598 381 1242 1052 1287 639 1413 378 524 8937

% Malignant postoperative by cytopathology diagnosis B

ATYP

FoN/HN

SUSP M

I

M

ND

18 4 N/A 8 2 7 10 N/A 2 5 N/A 6

0 N/A N/A 48 24 19 N/A 33 6 48 19 16

31 16 14 26 28 32 49 30 15 34 25 25

57 82 65 68 60 65 42 62 53 87 N/A 62

33 27 N/A 34 42 38 48 37 14 47 N/A 34

93 96 N/A 100 97 99 95 N/A 96 100 N/A 97

0 9 N/A 12 10 11 50 N/A 9 32 N/A 12

650

B. McIver / Oral Oncology 49 (2013) 645–653

logical diagnosis available to determine further management. The price of this approach is that thyroidectomy is often required purely for diagnostic purposes, something that has, until recently, been accepted as a ‘‘necessary evil’’. Nevertheless, it is a longstanding goal in the Endocrine and Surgical communities to improve the diagnostic capacity of FNA, and so avoid this situation. Molecular markers Remarkable advances have been made over the last two decades in our understanding of the underlying genetic and molecular changes that drive thyroid neoplasia.35 Chromosomal rearrangements involving the RET protooncogene, or the V600E point mutation in the BRAF gene, underpin the majority of PTC. Rearrangements of the PPARc gene have been implicated in a significant minority of follicular thyroid cancer (FTC), while point mutations of various RAS genes are found in both FTC and follicular adenomas (FA). It is unknown whether those FA that contain RASmutations are premalignant, but these mutations do seem to be growth promoting. In his groundbreaking work in the field of molecular diagnostics for thyroid cancer, Dr. Nikiforov developed the concept of a ‘‘mutation panel’’, performed on FNA material, to improve pre-operative diagnosis of thyroid nodules, particularly those with indeterminate cytology.36 That panel combines the results of somatic DNA mutation screening for BRAF V600E and 13 candidate RAS mutations, with mRNA testing for the two most common rearrangements in RET (RET/PTC 1 and 3) and the PAX8/PPARc rearrangement. Testing this approach on 214 samples classified cytologically as FN of Hurthle Cell Neoplasm (HN) allowed the correct pre-operative identification of 33 of the 58 histologically malignant nodules and 151 of the 156 histologically benign nodules, giving a sensitivity of 57% and specificity of 97%, and achieving a NPV of 86% (a 14% residual risk of malignancy in nodules that did not exhibit any of the tested mutations). The panel performed better in the 247 samples that made up the AUS/ FLUS cytology group, achieving a sensitivity of 63%, specificity of 99% and NPV of 94%. This residual 6% ROM, following negative mutation testing, is very close to the residual ROM in cytologically benign cytology and has led to recommendations that molecular markers could be used in this way to avoid unnecessary diagnostic surgery.36,7 It is important to point out, however, that fully 37% of the malignant nodules in the AUS/FLUS group were mutation negative, as were 43% of the cancers in the FN/HN group. Nevertheless, this panel has been adopted in various forms by a number of labs around the country and is commercially available through Asuragen Inc (Austin, TX) and Quest Diagnostics (Madison, NJ). An alternative approach has been developed by Veracyte Inc. (South San Francisco, CA), a venture capital funded biotech startup, and is being actively marketed under the trade name Afirma™. Based on the hypothesis that various types of thyroid nodules would have unique gene expression profiles, Veracyte set out to identify the expression signature of a benign nodule, in an effort to identify a subset of nodules that could safely be observed, so reducing the rate of ‘‘unnecessary’’ diagnostic surgery. Based on a commercially available gene expression chip (Affymetrix Inc, Santa Clara, CA), Veracyte has developed a ‘‘gene expression classifier’’ (GEC), using a proprietary algorithm to distinguish ‘‘benign’’ from ‘‘suspicious’’ nodules, based on the expression pattern of mRNA extracted from two dedicated FNA needle passes. That algorithm depends on an initial screening ‘‘cassette’’ of 23 genes designed to identify medullary thyroid cancer (MTC) and certain metastatic malignancies. Thereafter, the expression levels of 142 genes are processed through a ‘‘support vector machine’’ (SVM) that classifies the expression pattern as either ‘‘benign’’ or ‘‘suspicious’’.37 The SVM is a supervised machine learning algorithm that identifies patterns of gene expression in a recursive learning process, based

on a training set of mRNA samples derived from nodules with known histology. The initial training of the algorithm involved 315 samples, with mRNA derived from a combination of surgical tissue (178 samples) and FNA material (137 samples).37 The performance of the final GEC was assessed in a set of 577 FNA samples with indeterminate cytology, prospectively collected largely (though not completely) independent of the training set, and subsequently correlated with histopathology, read by three acknowledged experts.13 Cytology was not centrally evaluated in this study, but community cytology reports were reclassified into the most likely Bethesda classification. After a variety of exclusions, 265 of the 577 samples were included in the final analysis, of which 129 were classified as AUS/FLUS and 81 as FN/HN, with the remainder ‘‘suspicious for malignancy’’. Within the AUS/FLUS group, sensitivity of the GEC was 90%, specificity 53% and NPV 95%, while the figures for FN/HN were 90% sensitivity, 49% specificity and 94% NPV. The residual risk of malignancy in GEC-benign, cytologically AUS/FLUS or FN/HN nodule in this study was therefore 5–6%, again very close to that seen in reports of benign cytology. Consequently, Veracyte markets their Afirma GEC as a ‘‘rule out’’ test that can avoid the need for surgery. In their reported experience to date, over 50% of samples submitted to the GEC are reported as having a benign profile, leading Veracyte to claim that up to 50,000 patients could avoid surgery each year if the GEC was implemented across the country.38 Li et al. undertook a statistical modeling exercise, based on the initial cost of the Afirma GEC, and estimated probabilities and costs of surgery and management of its complications, and determined that the assay should be costeffective, with a cost-saving of $1400 per patient, with a modest improvement in quality of life.39 Our own data, presented so far only in abstract form (American Thyroid Association Annual Meeting, 2012, Quebec City, QE), are rather less compelling, however, with only 25% of samples receiving a benign GEC result. Of the samples that were GEC suspicious, and for whom final histology is known, only 15% have proven malignant, a PPV substantially lower than the 38% reported by Alexander et al. Furthermore, Li et al. in their cost effectiveness analysis, assumed that ‘‘almost three fourths of currently performed surgeries in patients with benign nodules’’ could be avoided. The much lower rate of surgery avoidance that could have been achieved by Alexander et al. (41% of calls were benign) and in our experience (25% benign) call into question that analysis. Additional real-world experience and independent assessment of Afirma GEC performance and cost-effectiveness is needed before this test is widely implemented. In order safely to avoid surgery, it is imperative that the residual risk of malignancy is known, following a benign call from any test. It is self evident that a physician would not fully trust a benign cytology report in a patient who had been exposed to radiation in childhood, had a growing thyroid nodule that was PET-positive and highly suspicious on ultrasound. Almost certainly, that patient would be recommended for a repeat biopsy, or even for a diagnostic thyroidectomy. The NCCN recommends observation ‘‘[i]f molecular testing predicts a risk of malignancy . . . approximately 5% or less’’.7 However, a similar level of skepticism needs to be applied to the results of molecular markers as most physicians already bring to the results of other tests. In statistical terms, the pre-test probability of malignancy influences the reliability of the test result, in calling either malignant or benign. Graphically, the relationship between the pre-test probability of malignancy and the achieved NPV is shown in Fig. 2A for the Afirma GEC (solid line) and for the Nikiforov oncogene panel (dashed line). If the pre-test probability of malignancy is sufficiently low, the NPV of both tests approach 100%, while at high ROM, the NPV falls towards zero. In the case of a cascaded test, performed only in the context of indeterminate cytopathology, the pre-test

B. McIver / Oral Oncology 49 (2013) 645–653

A

1

NPV

0.8 0.6 0.4 0.2 0

0

0.2

0.4

0.6

0.8

1

0.8

1

Prevalence of malignancy

B

1

PPV

0.8 0.6 0.4 0.2 0

0

0.2

0.4

0.6

Prevalence of malignancy Figure 2. (A) Negative predictive value (NPV) for Afirma-GEC and Nikiforov’s oncogene panel for AUS/FLUS, as a function of the pre-test probability of malignancy, based on sensitivity and specificity reported by Nikiforov et al. 201112 and Alexander et al. 2012.13 Afirma GEC has a reported sensitivity of 90% and specificity of 53%, while the Nikiforov panel has values of 63% and 99% respectively. As pre-test probability of malignancy (ROM) increases, NPV falls. Achieving a NPV of >95% requires a ROM of <23% for Afirma-GEC and <14% for the oncogene panel. (B) Positive predictive value (PPV) for Afirma-GEC (solid line) and Nikiforov’s oncogene panel for AUS/FLUS, as a function of pretest probability of malignancy, based on the same sensitivity and specificity values noted above.12,13 As ROM increases, PPV rises. The oncogene panel is a particularly powerful predictor of malignancy when ROM exceeds 10%.

ROM is determined by the cytopathology. Because the ROM in various cytological groups varies significantly between different centers (Table 3), the NPV of these tests will also vary significantly from center to center. If we accept the reported sensitivity and specificity of the Afirma GEC and the Nikiforov oncogene panel – an arguable position given the small size of the published series and the wide confidence intervals intrinsic to these numbers – the NPV of the two tests would vary dramatically, depending on where the cytology was read. For example, at Northwestern University, the risk of malignancy is 6% for AUS/FLUS and 15% for FN. Consequently, the NPV of Afirma GEC-benign would be 99% in AUS/FLUS and 97% in FN, while the NPV of a negative Nikiforov oncogene panel would be 98% in AUS/FLUS and 93% in FN. However, at Yale University, the risks for malignancy are 48% in AUS/ FLUS and 34% in FN, so the performance of Afirma GEC would decline to an NPV of only 90% in FN and 85% in AUS/FLUS, while the oncogene panel would give 81% and 74% NPV in these groups. Clearly, at Yale, no molecular testing should be applied to indeterminate cytology to avoid surgery, while at Nothwestern, either of the commercially available products would perform adequately for this purpose, at least in the AUS/FLUS group. As shown in Fig. 2B, the PPV of these tests similarly depends on the pre-test ROM, with the oncogene panel performing substantially better, because of its higher specificity. This high PPV may be useful if the goal is to determine the extent of surgery to be per-

651

formed, rather than address the question of whether surgery is necessary. With a mutation detected in one of the oncogenes, Nikiforov’s data suggest that a cancer surgery would likely be justified as an initial surgical step, if the pre-test ROM exceeds 10%. For mutation negative nodules, a more conservative procedure (lobectomy) might be considered. It is appropriate to inject one note of caution, however, by emphasizing that Nikiforov’s study was not fully blinded, so the pathology was read with the pathologist often aware of the mutation status of the nodule. This could have increased the measured specificity and so improve the apparent PPV of the test. Confirmatory studies and careful cost-effectiveness analysis are again necessary and appropriate before the oncogene panel can be recommended for widespread use in determining surgical extent. The specific niche suggested for the Afirma GEC is to determine the need for surgery. In this situation, the only appropriate use of the GEC would be in centers that can achieve a pre-test ROM of <23%, to allow the 95% NPV to be attained. However, most practices must necessarily use cytology services whose precise ROM is unknown and may well be higher than this within what they define as AUS/FLUS or FN/HN. This includes all of the major commercial pathology labs, including Thyroid Cytology Partners (TCPs, Austin, TX), the group that Veracyte requires physicians to use in order to gain access to their GEC assay. Because the ROM is unknown, the true NPV of the Afirma GEC, as commercially available, is also unknown, undermining the primary reason to order the test. Unless cytology is performed in a center that can assess its performance on an ongoing basis, with well defined ROM within its various cytologic categories, Afirma should be applied with caution. It is inappropriate – and misleading – to claim that the Afirma GEC can offer a defined NPV, or that surgery can be safely avoided by use of this or any other test, without a clearly stated pre-test ROM. Consequently, Afirma cannot, in its currently marketed form, meet the criteria established by the NCCN in its guidelines, that observation in lieu of surgery be considered if molecular testing predicts a risk of malignancy of approximately 5% or less.

Management of thyroid nodules Benign nodules are usually considered appropriate for observation, unless those nodules are large, compressive, or clinically concerning, or if the patient chooses removal for reasons of cosmesis or psychological concern. As noted previously, a small proportion of these nodules – some adenomatous nodules and many nodules that arise in the context of Hashimoto’s thyroiditis – might respond to mild TSH suppression. Despite the conflicting nature of the data, caused in part by the failure of most studies to take into account the pathology of the nodule, there is little doubt that at least a proportion of these nodules shrink with thyroid hormone suppression. A rigorous meta-analysis by the NIH group showed convincing evidence of a small but significant benefit in the regression of benign solitary thyroid nodules when the results of the available five randomized controlled trials were combined (Fig. 3).40 Overall, they found a 2-fold higher likelihood of achieving a >50% reduction in nodule size following thyroid hormone suppression compared with placebo control. A large simple clinical trial of appropriately selected adenomatous or regenerative nodules is clearly required to answer this question definitively. In the meantime, as stated by Csako et al. ‘‘[the] effect is biologically plausible, and is supported by both experimental evidence and expert opinion’’. There are, however, adverse consequences of thyroid hormone suppression, so any attempt to shrink a benign thyroid nodule should ensure that the patient is not harmed by the treatment. The dose of thyroid hormone should be tailored to achieve a TSH at, or marginally below the lower limit of normal, while maintain-

652

B. McIver / Oral Oncology 49 (2013) 645–653

Figure 3. Benefit ratios for the effect of thyroid hormone suppression therapy on the regression of benign solitary thyroid nodules in randomized control trials. (Numbers refer to control/treated patients), demonstrating the overall small but significant benefit from treatment with thyroid hormone suppression. From: Yousef et al. 2010.40

ing thyroid hormone concentrations within their normal ranges. Such subclinical thyrotoxicosis minimizes the risk of symptoms, but there remain concerns about its impact on the development of osteoporosis and the risk of atrial fibrillation.41 For nodules classified as AUS/FLUS, the usual recommendation has been to consider repeating the FNA. This makes sense if the cause of the AUS classification is the presence of clotting artifact or other cause of poor quality smears. However, other causes of this categorization are less likely to result in a different outcome. For both AUS/FLUS and FN/HN, perhaps the best, and certainly the least expensive, tool is to refer the samples to a known reference center for a cytological second opinion. If AUS/FLUS or FN/ HN is confirmed, then repeat cytology should be considered, including – under carefully controlled circumstances – the use of molecular markers, if that center has a known and sufficiently low ROM in AUS/FLUS and FN that the NPV can reasonably be expected to meet the 5% threshold. Otherwise, these nodules are, on average, more appropriately managed by surgical resection. The extent of necessary surgery could then usefully be influenced by molecular markers, specifically the Nikiforov panel, whose high specificity and consequently high PPV could minimize the likelihood that a later completion thyroidectomy proves necessary. Cost effectiveness of this approach has been claimed, though the actual costs will again be influenced by the underlying ROM, which determines PPV, just as it does NPV. With low ROM, the PPV will be lower and the costs higher, whereas high ROM could improve cost-effectiveness of the molecular testing. It seems unlikely that the PPV of the Afirma GEC will be sufficient to influence surgical extent in a cost-effective manner, thought that has not yet been subject to formal evaluation. Nodules that are suspicious for malignancy have a ROM that ranges from 65% to 85% and are optimally managed by surgery, typically a lobectomy or total thyroidectomy, as discussed elsewhere in this volume. The diagnosis of a malignant nodule, of course, also mandates appropriate surgery. Conclusions The discovery of a thyroid nodule should lead to an appropriate work up, which includes clinical assessment, measurement of TSH,

ultrasound evaluation and FNA cytology in appropriate cases. If the nodule proves to be benign, and if the patient and physician wish to treat it, than a careful assessment of the underlying pathology is essential. Patients with autonomously functioning thyroid nodules, or toxic multi-nodular goiter, can safely be treated with either radioactive iodine or surgery. In patients with Hashimoto’s disease and a regenerative nodule, a carefully supervised trial of thyroxine therapy, may prove helpful and relieve both patient and physician of considerable anxiety, while avoiding an unnecessary surgical procedure in a significant proportion of cases. Nodules that carry a high probability of malignancy (suspicious for malignancy or malignant) should be referred for appropriate surgery. For indeterminate cytology, molecular markers are likely to play an increasing role in the future, but at this time, their application may be premature in most clinical setting, except when a well-defined ROM exists, so that both NPV and PPV are defined and clinical decisions can be made confidently. Widespread use of molecular markers, however, is not yet justified on the basis of the data so far available, particularly when those markers target the goal of reducing diagnostic surgery. There is a clear and present danger that use of this approach could leave a significant number of patients undertreated for a malignancy that, if caught early and managed appropriately, carries only minimal risk of mortality, but which, if allowed to progress unchecked, could become significantly threatening.

Conflict of interest statement Dr McIver has received honoraria for speaking engagements and scientific consulting from Veracyte Inc and from Asuragen Inc, manufacturers of the two commercially available molecular assays marketed to assist in the evaluation of thyroid nodules.

References [1] Mazzaferri EL. Management of a solitary thyroid nodule. N Engl J Med 1993;328:553–9.

B. McIver / Oral Oncology 49 (2013) 645–653 [2] American Cancer Society 2012 Cancer Facts & Figures. Atlanta: American Cancer Society. [accessed 01.02.13]. [3] Caruso DR, Mazzaferri EL. Fine needle aspiration biopsy in the management of thyroid nodules. Endocrinologist 1991;1:194–202. [4] Yassa L, Cibas ES, Benson CB, Frates MC, Doubilet PM, Gawande AA, et al. Longterm assessment of a multidisciplinary approach to thyroid nodule diagnostic evaluation. Cancer 2007;111:508–16. [5] Ross DS. Predicting thyroid malignancy. J Clin Endocrinol Metab 2006;91:4253–5. [6] Gharib H. Single thyroid nodule. Curr Ther Endocrinol Metab 1997;6:112–7. [7] Tuttle et al. Thyroid carcinoma. J Natl Compr Cancer Network 2010;8:1228–74. [8] Cooper DS, Doherty GM, Haugen BR, Hauger BR, Kloos RT, Lee SL, et al. Revised American Thyroid Association management guidelines for patients with thyroid nodules and differentiated thyroid cancer. Thyroid 2009;19:1167–214. [9] Gharib H, Papini E, Paschke R, Duick DS, Valcavi R, Hegedüs L, et al. American Association of Clinical Endocrinologists, Associazione Medicie Endocrinologi, and European Thyroid Association medical guidelines for clinical practice for the diagnosis and management of thyroid nodules: executive summary of recommendations. J Endocrinol Invest 2010;33:51–6. [10] Fagin J, Mitsiades N. Molecular pathology of thyroid cancer: diagnosis and clinical implications. Best Prac Res Clin Endocrinol Metab 2008;22:955–69. [11] Nikiforova MN, Nikiforov YE. Molecular genetics of thyroid cancer: implications for diagnosis, treatment and prognosis. Expert Rev Mol Diagn 2008;8:83–95. [12] Nikiforov YE et al. Impact of mutational testing on the diagnosis and management of patients with cytologically indeterminate thyroid nodules: a prospective analysis of 1056 FNA samples. J Clin Endocrinol Metab 2011;96:3390–7. [13] Alexander EK et al. Preoperative diagnosis of benign thyroid nodules with indeterminate cytology. New Eng J Med 2012;367:705–15. [14] Fadda G, LiVolsi VA. Histology and aspiration cytology of benign thyroid diseases. Rays 1999;24:182–96. [15] Yang GC, Stern CM, Messina AV. Cystic papillary thyroid carcinoma in fine needle aspiration may represent a subset of the encapsulated variant in WHO classification. Diagn Cytopathol 2010;38:721–6. [16] Sung JY, Kim YS, Choi H, Lee JH, Baek JH. Optimum first-line treatment technique for benign cystic thyroid nodules: ethanol ablation or radiofrequency ablation? Am J Roentgenol 2011;196. W210-4. [17] Paschke R. Molecular pathogenesis of nodular goiter. Langenbecks Arch Surg 2011;396:1127–36. [18] Brito JP, Yanur AJ, Prokop LJ, McIver B, Murad MH, Montori V. Prevalence of thyroid cancer in multinodular goiter vs. single nodule: a systemic review and meta-analysis. Thyroid, October 2012. ePub ahead of print. [19] Krohn K, Emmrich P, Ott N, Paschke R. Increased thyroid epithelial cell proliferation in toxic thyroid nodules. Thyroid 1999;9:241–6. [20] Gharib H, Mazzaferri EL. Thyroxine suppressive therapy in patients with nodular thyroid disease. Ann Int Med 1998;128:386–94. [21] Gharib H. Single thyroid nodule. Curr Ther Endocrinol Metab 1997;6:112–7. [22] Boelaert K. The association between serum TSH concentration and thyroid cancer. Endocr Relat Cancer 2009;16:1065–72. [23] Pacini F, Schlumberger M, Dralle H, Elisei R, Smit JW, Wiersinga W, et al. European consensus for the management of patients with differentiated thyroid carcinoma of the follicular epithelium. Eur J Endocrinol 2006;154:787–803.

653

[24] Seiberling KA, Dutra JC, Gunn J. Ultrasound-guided fine needle aspiration biopsy of thyroid nodules performed in the office. Laryngoscope 2008;118:228–31. [25] Nóbrega LH, Paiva FJ, Nóbrega ML, Mello LE, Fonseca HA, Costa SO, et al. Predicting malignant involvement in a thyroid nodule: role of ultrasonography. Endocr Pract 2007;13:219–24. [26] Liu YI, Kamaya A, Desser TS, Rubin DL. A bayesian network for differentiating benign from malignant thyroid nodules using sonographic and demographic features. Am J Roentgenol 2011;196:W598–605. [27] Bhatia K, Tong CS, Cho CC, Yuen EH, Lee J, Ahuja AT. Reliability of shear wave ultrasound elastography for neck lesions identified in routine clinical practice. Ultraschall Med 2012;33:463–8. [28] Hsiao YC, Wu PS, Chiu NT, Yao WJ, Lee BF, Peng SL. The use of dual-phase 18FFDG PET in characterizing thyroid incidentalomas. Clin Radiol 2011;66:1197–202. [29] Kim MJ, Kim EK, Park SI, Kim BM, Kwak JY, Kim SJ, et al. US-guided fine-needle aspiration of thyroid nodules: indications, techniques, results. Radiographics 2008;28:1869–86. [30] Wu M. A comparative study of 200 head and neck FNAs performed by a cytopathologist with versus without ultrasound guidance: evidence for improved diagnostic value with ultrasound guidance. Diagn Cytopathol 2011;39:743–51. [31] Bisi H, de Camargo RY, Longatto Filho A. Role of fine-needle aspiration cytology in the management of thyroid nodules: review of experience with 1,925 cases. Diagn Cytopathol 1992;8:504–10. [32] Cibas ES, Syed AZ. NCI thyroid FNA state of the science conference. The Bethesda system for reporting thyroid cytopathology. Am J Clin Pathol 2009;132:658–65. [33] Lobo C, McQueen A, Beale T, Kocjan G. The UK royal college of pathologists thyroid fine-needle aspiration diagnostic classification is a robust tool for the clinical management of abnormal thyroid nodules. Acta Cytol 2011;55:499–506. [34] Wang CC, Friedman L, Kennedy GC, Wang H, Kebebew E, Steward DL, et al. A large multicenter correlation study of thyroid nodule cytopathology and histopathology. Thyroid 2011;21:243–51. [35] Fagin JA. The Jeremiah Metzger Lecture: intelligent design of cancer therapy: trials and tribulations. Trans Am Clin Climatol Assoc 2007;118:253–61. [36] Nikiforova MN, Nikiforov YE. Molecular diagnostics and predictors in thyroid cancer. Thyroid 2009;19:1351–61. [37] Chudova D, Wilde JI, Wang ET, Wang H, Rabbee N, Egidio CM, et al. Molecular classification of thyroid nodules using high-dimensionality genomic data. J Clin Endocrinol Metab 2010;95:5296–304. [38] Duick DS, Klopper JP, Diggans JC, Friedman L, Kennedy GC, Lanman RB, et al. The impact of benign gene expression classifier test results on the endocrinologistpatient decision to operate on patients with thyroid nodules with indeterminate fine-needle aspiration cytopathology. Thyroid 2012;22:996–1001. [39] Li H, Robinson KA, Anton B, Saldanha IJ, Ladenson PW. Cost-effectiveness of a novel molecular test for cytologically indeterminate thyroid nodules. J Clin Endocrinol Metab 2011;96:E1719–26. [40] Yousef A, Clark J, Doi SA. Thyroxine suppression therapy for benign nonfunctioning solitary thyroid nodules: a quality-effects meta-analysis.. Clin Med Res 2010;8:150–8. [41] Sawin CT, Geller A, Wolf PA, et al. Low serum thyrotropin concentrations as a risk factor for atrial fibrillation in older persons. N Engl J Med 1994;331:1249–52.